The "Process" of John Edward

[Lurker]

Unfortunately, the chi-square test results become suspect for cells with expected frequencies below 5. The results are also suspect for total chi-square values under 50.

Bill is correct here. From my memory, a cell with an expected frequency below 5 might be acceptable if the others were above 5. That all but two are below 5 bodes ill for the chi-square test's accuracy.

Lurker

 

[BNiles]

The table of the counts, per transcript and the total counts, is shown below:

(9/10/01, 9/6/02, 3/28/03, Observed total counts)
A: (0, 1, 0, 1)
C: (2, 0, 2, 4)
D: (0, 1, 0, 1)
J: (4, 2, 2, 8)
M: (0, 1, 1, 2)
R: (1, 1, 1, 3)

...

Expected counts = 80 guesses*1990 combined adjusted male and female Census Bureau data frequencies =
A: 5.18
C: 5.768
D: 5.936
J: 10.68
M: 8.030
R: 6.40

I'm not a statistician either, but IMO a larger sample size may have been more revealing. What I found interesting about this analysis wasn't the frequency of any one letter, but rather the size commonality of frequencies throughout. If we place the letters in order of the frequency sizes the lists are only off by two...ie:

Test_____Letter
8__________J
4__________C
3__________R
2__________M
1__________D
1__________A

Expect_____Letter
10.68_______J
8.030_______M
6.400_______R
5.936_______D
5.768_______C
5.180_______A

I think that with a larger sample size, we'll see more telling results.

 

[T'ai Chi]

Bill is correct here. From my memory, a cell with an expected frequency below 5 might be acceptable if the others were above 5.

Lurker

That is entirely correct, Lurker and Bill.

The point is that it seems better, in my opinion, to analyze a set of high frequency letters. Otherwise, you are subjectively choosing one high frequency letter for the analysis.

As we saw, J and C were used more than expected, but A, D, M, R were used less than expected. It is hard to decide how to interpret that in my opinion.

 

[neofight]

Well, in "Blues Clues" there are characters; Mr. Pepper, Mrs. Salt and they have two children, Cinnamon and Paprika. Maybe John Edward is getting these characters in his visions? Is John Edward a big fan of "Blues Clues"?

LOL Lurker. You might be kidding, I don't know, but you are actually quite right in what you say here. Now it's true that John's little boy is only turning one this month, but soon enough he will be watching "Blues Clues", the "Wiggles", Dora the Explorer, Spongebob Squarepants, and Sesame Street etc. with him, and his personal frame of reference in kid's pop culture will expand greatly.

I love it when the dads get all involved with their kids. .....neo

 

[CFLarsen]

I think we are forgetting something. When JE guesses names, he shouldn't take from the whole pool of names. He should take from the most common names.

Which he does.

I looked at the 50 and 100 most common names, both male and female and took the initials only. Guess what?

==================
Top 100 Letters
==================
12_________J
10_________D
10_________M
9__________S
8__________A
8__________C
8__________R


==================
Top 200 Letters
==================
30_________J
18_________A
16_________R
14_________C
14_________S
13_________D
13_________M


==================
JE's guesses
==================
8__________J
4__________C
3__________R
2__________M
1__________D
1__________A


J is the most popular by far in both tallies. All 6 of JE's most common initials are among the top 7 most common initials among the 100 and 200 most popular names.

This approach makes perfectly sense, if we only look at the most popular names: After all, that's where the possibilities are greatest.

Don't tell me he isn't playing the odds here. I know, small numbers. But it stinks....

 

[Garrette]

Okay, I'm going to flop to the JE supporter side for a moment, for sake of argument. Actually, I think the point I'm about to bring up is valid, if not brought up already.

Claus, your numbers appear on the surface to be damning.

However, if JE uses initials according to the frequency of most used initials, mightn't it be because you would expect the spirits, the sitters, and their extended circle of acquaintances to follow the same distribution.

I don't know if I'm saying this well...

If the most common initials are JDMSACR, then why should we expect a bunch of spirits to come through with Zs?

Someone please destroy this argument for me...

 

[Ersby]

As we saw, J and C were used more than expected, but A, D, M, R were used less than expected. It is hard to decide how to interpret that in my opinion.

R was used exactly as expected (it's hard for JE to make 0.1 of a guess on a live 'phone-in). D is used less, but that's due to one of JE favoured tactics. A and M are barely one away from their expected frequency, so it could be that the number of guesses is simply too small.

Personally, I think Garratte's right. The guesses expected by a medium should be similar to the distribution of names in the general population. As should the guesses of a cold reader. (Both, of course, throwing in a few rarer guesses for show.)

I'm going to dig through some of the seminar notes on tvtalkshows, and do a name count with those. See what turns up.

 

[CFLarsen]

Garrette,

On the surface, you are right.

Look at what JE is doing: He is not gue...throwing out names, he is gue...throwing out initials. That widens the field considerably - he only has to pick among the top 6 or 7 letters to get a 65% chance of a hit if he takes the first 100 most common initials.

Consider if he gue...threw out names: Far less chance of a hit.

So, why doesn't he throw out names? He can hear full words, terms, he can get clear images, whatever. But not names.

That is what is damning: In everything else, he can get full words. Not names. There, he has to play the game of most common initials. He can do it, because his fans lap it up.

It's all a matter of inflating the hits and get away with it.

 

[Garrette]

Oh, I understand that, Claus. I have no doubt that JE is a performer and not a medium (sorry, neo and Clancie, but there ya have it).

But there is an analysis in progress. The purpose of the analysis is to demonstrate, somewhat scientifically, that JE is or isn't acting in accordance with a cold reader (or somesuch hypothesis or null hypothesis).

My point is that the analysis, insofar as it has been to date, can be of no use in disproving JE's mediumistic abilities.

I guess this was sort of my point in my initial questions many posts ago.

An analysis of the distribution of initials is insufficient without a concurrent analysis of their success.

If over the course of x tapings, y number of Js would be expected to result, by chance, in z hits, and in fact do result in z hits, then we've learned that JE doesn't operate above chance.

But if the same y number of Js result in significantly more than z hits, then there is evidence in support of his claims.

Simply proving that JE tosses out initials in a distribution equal to their distribution among the most popular names does nothing, meine meinung.

 

[T'ai Chi]

He should take from the most common names.

Which he does.

I personally don't think that is in dispute. Afterall, common names here might be common names in the afterlife (assuming, of course, there is one).

Don't tell me he isn't playing the odds here.

Well, that is why we have to find statistical evidence, so we can say with good probability that he is (or isn't).

 

[CFLarsen]

Garrette,

I agree that we cannot look at just one aspect. We have to look at what is happening and try to find a rational explanation. Spin many theories, like Sagan said.

After all, cold reading is made up of many techniques and strategies.....

 

[Ersby]

Okay, I did the count, and scoring them in the same way as above, the hypothesis as presented here, suffers quite a knock.

Number of guesses concerning names: 77

Looking at those letters highlighted for analysis, we find:

(frequency____expected frequency)
A: 4_____4.9
C: 7_____5.5
D: 5_____5.7
J: 12_____10.2
M: 5_____7.7
R: 4______6.1

So no real change in tactic, except for a lesser use of R.

So, for completeness sake (though it won’t change much) here’s the total for LKL and seminar notes combined:

A: 6______8.3
C: 13_____9.3
D: 7______9.5
J: 26______17.2
M: 9______12.9
R: 8_______10.3

At first glance it looks like, of the top six, JE only overuses two: C and J. But, as I said, D often seems to be “held back” for readings in which children are mentioned, plus these figures don’t allow for the letter B (8 guesses over all transcripts), which as well as the genuine B names, can also be stretched to include William (Billy), Robert (Bobby) and Elizabeth (Betty).

JE does have a curious propensity to guess L names. His favourite seems to be “Lou, Louis, Louise”, and in fact in one reading after getting a hit with it he says it “never lets him down”. Make of that what you will!

And as I said earlier, the circumstances of him making the guess need to be known. I couldn’t say for sure, but while I was counting up, the C and M guesses seemed to be slanted towards females. Of course, JE’s audience is mostly females, so it’s a shaky hypothesis at best. I’ll take a closer look later perhaps, but time’s pressing so I’ll wrap up here.

 

[T'ai Chi]

I was wondering if we could think of anything else besides letter counts that we could analyze in a statistical manner.

One thing that I was interested in a while back was in counting the number of questions JE asks per reading. Not clarification questions like "Do you understand that?", or filler questions, but questions more directly relevant to the reading, like "Who is Betty?", etc., for example.

I was hypothesizing that perhaps JE needs to use a certain number of questions to have a "successful" reading.

Is JE's reading success dependent on the number of questions he asks? Should it be if he gets his messges from the spirits? -Or can he have a successful reading with very few or no questions?

Perhaps someone could go through each reading and score it a 1 for "success" and a 0 for "failure" (of course that could be very subjective). Someone could also go through and count the number of directly relevant questions JE asks (this is somewhat subjective too) for that reading. So for each reading, we'd have two numbers in the form:

(# of directly relevant questions, a 0 or 1).

I think then, at least for starters, one could see if the success group's average number of questions is significantly different from the failure groups' average number of questions.

I would think that the average of success group being significantly greater than the average of the failure group, could be considered evidence of cold reading.

This idea popped into my head a while back. I'm not entirely sure if this would be a worthwhile analysis though, or what it would really mean (it is late here too! ). The analysis could also be very sensitive to the raters' somewhat subjective ratings.

 

[T'ai Chi]

Perhaps we could also keep a continually updated list of JE's letter frequencies.

That is, we could have a list like:

JE's Letter Frequencies

A: p_A
B: p_B
C: p_C
D: p_D
.
.
.
Z: p_Z
(where p_A is the frequency for the letter A, for example)

One could simply (but laboriously) build this list from his past readings from a variety of sources. This list could simply be continually added to as JE does more readings.

We could simply use it as a descriptive tool to keep track of which letters JE uses. We could also use it as a long-run tool to use for inference to compare his frequencies to the Census Bureau frequencies sometime in the future when the lists gets built up.

We could do this with other mediums and see if their lists are similar or not.

Just another idea.

 

[Lurker]

.

However, if JE uses initials according to the frequency of most used initials, mightn't it be because you would expect the spirits, the sitters, and their extended circle of acquaintances to follow the same distribution.

Someone please destroy this argument for me...

Garrete:

I think you are misunderstanding the problem. We were looking at. We were interested to determine if JE is using common letters MORE than the census distribution. The line of thought was that if JE's letter distribution follows the census then it would NOT be an argument in favor of him cold reading. If he uses common letters disproportionately MORE than the census figures then the idea was that this would be evidence that he is cold reading. (Zeroing in on letters he KNOWS to be more common)

Now, the analysis seems suspect as counting methods play a large role in our data. Further, I am not sure the the basic logic of (IF A THEN B) is very good either.

But it was a fun exercise.

Lurker

 

[Garrette]

Originally Posted by Lurker
Garrete:

I think you are misunderstanding the problem.

It certainly wouldn't be the first time...

Anyway, I understand your explanation. Thanks.

It is an interesting exercise, but I still question the value of it.

Seems we're doing something like: "Let's try something that, if it gives one result, will merely indicate the possibility of one explanation, but if it gives any other result, doesn't disprove anything at all."

Actually, I only question the value of it alone; if we now compared the initials guesses with the hits for the same initials, we'd be getting somewhere. I think. I wasn't a statistician yesterday, and I think I'm still not one today. But I've read a lot of Tai Chi's posts lately so maybe I am...

 

[BNiles]

I think that T'ai Chi is on to something here.

IMO, it has been demonstrated that JE uses the letter "J" well above the expected frequency. Take from that what you will. I also agree that this analysis only looks at one aspect of the overall process involved in cold reading. T'ai Chi, along with several others, has recommended other sampling that would yield telling data, but it requires more research than a casual poster is willing to compile.

Here's where the skeptic runs into trouble.

If a skeptic argues with only one analysis, absolute proof cannot be obtained = JE wins.
A skeptic won't attempt to do all analyses, because the outcome doesn't net enough profit to warrant the expansive workload involved = JE wins.

Maybe this thread should be on a paradox board

What came first, the spirit or the medium?

 

[CFLarsen]

If he uses common letters disproportionately MORE than the census figures then the idea was that this would be evidence that he is cold reading. (Zeroing in on letters he KNOWS to be more common)

Then there is evidence that he is cold reading. He chooses the letter J twice as often as the second letter. Among the 100 most common names (a pretty big chunk), the letter J is only slightly more common than D and M (and S).

I know, I know...small sample. However, we should keep an eye out for more readings.

 

[Lurker]

Then there is evidence that he is cold reading. He chooses the letter J twice as often as the second letter. Among the 100 most common names (a pretty big chunk), the letter J is only slightly more common than D and M (and S).

I know, I know...small sample. However, we should keep an eye out for more readings.

But even this conclusion is somewhat weak in my opinion. If we assume JE is cold reading and using the census figures to his advantage we would be assuming;

1. He is knowledgable about census figures
2. That he is using a letter strategy having broken the census data into first letter components
3. He hears all names and letters with equal certainty. Perhaps he hears his own letter "J" for John better than others. Or perhaps he SEES it better as he has better symbology associated with the letter "J"
4. We would expect other letters with similar frequencies as "J" (Such as D and M as you mentioned above) to be used disproportionately high as well. Is this the case? Why would he just choose "J" to use more often?

Well, those are the only ones I thought of right now. You see how WE are basing our conclusion on certain assumptions that may or may not be valid?

Lurker

 

[Lurker]

And don't get me wrong, I agree it is evidence. But the QUALITY of the evidence is debateable.

Lurker